| ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py |
| ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx -O3 | FileCheck %s --check-prefixes=CHECK,AVX,AVX-32 |
| ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx -O3 | FileCheck %s --check-prefixes=CHECK,AVX,AVX-64 |
| ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512f -O3 | FileCheck %s --check-prefixes=CHECK,AVX512F,AVX512F-32 |
| ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f -O3 | FileCheck %s --check-prefixes=CHECK,AVX512F,AVX512F-64 |
| ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512vl -O3 | FileCheck %s --check-prefixes=CHECK,AVX512VL,AVX512VL-32 |
| ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl -O3 | FileCheck %s --check-prefixes=CHECK,AVX512VL,AVX512VL-64 |
| ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=avx512dq -O3 | FileCheck %s --check-prefixes=CHECK,AVX512DQ |
| ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx512dq -O3 | FileCheck %s --check-prefixes=CHECK,AVX512DQ |
| ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=avx512dq,avx512vl -O3 | FileCheck %s --check-prefixes=CHECK,AVX512DQVL |
| ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx512dq,avx512vl -O3 | FileCheck %s --check-prefixes=CHECK,AVX512DQVL |
| |
| |
| declare <4 x i64> @llvm.experimental.constrained.fptosi.v4i64.v4f64(<4 x double>, metadata) |
| declare <4 x i64> @llvm.experimental.constrained.fptoui.v4i64.v4f64(<4 x double>, metadata) |
| declare <4 x i64> @llvm.experimental.constrained.fptosi.v4i64.v4f32(<4 x float>, metadata) |
| declare <4 x i64> @llvm.experimental.constrained.fptoui.v4i64.v4f32(<4 x float>, metadata) |
| declare <4 x i32> @llvm.experimental.constrained.fptosi.v4i32.v4f64(<4 x double>, metadata) |
| declare <4 x i32> @llvm.experimental.constrained.fptoui.v4i32.v4f64(<4 x double>, metadata) |
| declare <4 x i16> @llvm.experimental.constrained.fptosi.v4i16.v4f64(<4 x double>, metadata) |
| declare <4 x i16> @llvm.experimental.constrained.fptoui.v4i16.v4f64(<4 x double>, metadata) |
| declare <4 x i8> @llvm.experimental.constrained.fptosi.v4i8.v4f64(<4 x double>, metadata) |
| declare <4 x i8> @llvm.experimental.constrained.fptoui.v4i8.v4f64(<4 x double>, metadata) |
| declare <4 x i8> @llvm.experimental.constrained.fptosi.v4i8.v4f32(<4 x float>, metadata) |
| declare <4 x i8> @llvm.experimental.constrained.fptoui.v4i8.v4f32(<4 x float>, metadata) |
| declare <4 x i1> @llvm.experimental.constrained.fptosi.v4i1.v4f64(<4 x double>, metadata) |
| declare <4 x i1> @llvm.experimental.constrained.fptoui.v4i1.v4f64(<4 x double>, metadata) |
| declare <8 x i32> @llvm.experimental.constrained.fptosi.v8i32.v8f32(<8 x float>, metadata) |
| declare <8 x i32> @llvm.experimental.constrained.fptoui.v8i32.v8f32(<8 x float>, metadata) |
| declare <8 x i16> @llvm.experimental.constrained.fptosi.v8i16.v8f32(<8 x float>, metadata) |
| declare <8 x i16> @llvm.experimental.constrained.fptoui.v8i16.v8f32(<8 x float>, metadata) |
| declare <8 x i8> @llvm.experimental.constrained.fptosi.v8i8.v8f32(<8 x float>, metadata) |
| declare <8 x i8> @llvm.experimental.constrained.fptoui.v8i8.v8f32(<8 x float>, metadata) |
| declare <8 x i1> @llvm.experimental.constrained.fptosi.v8i1.v8f32(<8 x float>, metadata) |
| declare <8 x i1> @llvm.experimental.constrained.fptoui.v8i1.v8f32(<8 x float>, metadata) |
| |
| define <4 x i64> @strict_vector_fptosi_v4f64_to_v4i64(<4 x double> %a) #0 { |
| ; AVX-32-LABEL: strict_vector_fptosi_v4f64_to_v4i64: |
| ; AVX-32: # %bb.0: |
| ; AVX-32-NEXT: pushl %ebp |
| ; AVX-32-NEXT: .cfi_def_cfa_offset 8 |
| ; AVX-32-NEXT: .cfi_offset %ebp, -8 |
| ; AVX-32-NEXT: movl %esp, %ebp |
| ; AVX-32-NEXT: .cfi_def_cfa_register %ebp |
| ; AVX-32-NEXT: andl $-8, %esp |
| ; AVX-32-NEXT: subl $32, %esp |
| ; AVX-32-NEXT: vmovlps %xmm0, {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: vmovhps %xmm0, {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: vextractf128 $1, %ymm0, %xmm0 |
| ; AVX-32-NEXT: vmovlps %xmm0, {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: vmovhps %xmm0, (%esp) |
| ; AVX-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fldl (%esp) |
| ; AVX-32-NEXT: fisttpll (%esp) |
| ; AVX-32-NEXT: wait |
| ; AVX-32-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; AVX-32-NEXT: vpinsrd $1, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX-32-NEXT: vpinsrd $2, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX-32-NEXT: vpinsrd $3, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX-32-NEXT: vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX-32-NEXT: vpinsrd $1, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX-32-NEXT: vpinsrd $2, (%esp), %xmm1, %xmm1 |
| ; AVX-32-NEXT: vpinsrd $3, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX-32-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX-32-NEXT: movl %ebp, %esp |
| ; AVX-32-NEXT: popl %ebp |
| ; AVX-32-NEXT: .cfi_def_cfa %esp, 4 |
| ; AVX-32-NEXT: retl |
| ; |
| ; AVX-64-LABEL: strict_vector_fptosi_v4f64_to_v4i64: |
| ; AVX-64: # %bb.0: |
| ; AVX-64-NEXT: vextractf128 $1, %ymm0, %xmm1 |
| ; AVX-64-NEXT: vcvttsd2si %xmm1, %rax |
| ; AVX-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX-64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0] |
| ; AVX-64-NEXT: vcvttsd2si %xmm1, %rax |
| ; AVX-64-NEXT: vmovq %rax, %xmm1 |
| ; AVX-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0] |
| ; AVX-64-NEXT: vcvttsd2si %xmm0, %rax |
| ; AVX-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] |
| ; AVX-64-NEXT: vcvttsd2si %xmm0, %rax |
| ; AVX-64-NEXT: vmovq %rax, %xmm0 |
| ; AVX-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0] |
| ; AVX-64-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX-64-NEXT: retq |
| ; |
| ; AVX512F-32-LABEL: strict_vector_fptosi_v4f64_to_v4i64: |
| ; AVX512F-32: # %bb.0: |
| ; AVX512F-32-NEXT: pushl %ebp |
| ; AVX512F-32-NEXT: .cfi_def_cfa_offset 8 |
| ; AVX512F-32-NEXT: .cfi_offset %ebp, -8 |
| ; AVX512F-32-NEXT: movl %esp, %ebp |
| ; AVX512F-32-NEXT: .cfi_def_cfa_register %ebp |
| ; AVX512F-32-NEXT: andl $-8, %esp |
| ; AVX512F-32-NEXT: subl $32, %esp |
| ; AVX512F-32-NEXT: vmovlps %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: vmovhps %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: vextractf128 $1, %ymm0, %xmm0 |
| ; AVX512F-32-NEXT: vmovlps %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: vmovhps %xmm0, (%esp) |
| ; AVX512F-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fldl (%esp) |
| ; AVX512F-32-NEXT: fisttpll (%esp) |
| ; AVX512F-32-NEXT: wait |
| ; AVX512F-32-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; AVX512F-32-NEXT: vpinsrd $1, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX512F-32-NEXT: vpinsrd $2, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX512F-32-NEXT: vpinsrd $3, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX512F-32-NEXT: vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX512F-32-NEXT: vpinsrd $1, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX512F-32-NEXT: vpinsrd $2, (%esp), %xmm1, %xmm1 |
| ; AVX512F-32-NEXT: vpinsrd $3, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX512F-32-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX512F-32-NEXT: movl %ebp, %esp |
| ; AVX512F-32-NEXT: popl %ebp |
| ; AVX512F-32-NEXT: .cfi_def_cfa %esp, 4 |
| ; AVX512F-32-NEXT: retl |
| ; |
| ; AVX512F-64-LABEL: strict_vector_fptosi_v4f64_to_v4i64: |
| ; AVX512F-64: # %bb.0: |
| ; AVX512F-64-NEXT: vextractf128 $1, %ymm0, %xmm1 |
| ; AVX512F-64-NEXT: vcvttsd2si %xmm1, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512F-64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0] |
| ; AVX512F-64-NEXT: vcvttsd2si %xmm1, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm1 |
| ; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0] |
| ; AVX512F-64-NEXT: vcvttsd2si %xmm0, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512F-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] |
| ; AVX512F-64-NEXT: vcvttsd2si %xmm0, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm0 |
| ; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0] |
| ; AVX512F-64-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX512F-64-NEXT: retq |
| ; |
| ; AVX512VL-32-LABEL: strict_vector_fptosi_v4f64_to_v4i64: |
| ; AVX512VL-32: # %bb.0: |
| ; AVX512VL-32-NEXT: pushl %ebp |
| ; AVX512VL-32-NEXT: .cfi_def_cfa_offset 8 |
| ; AVX512VL-32-NEXT: .cfi_offset %ebp, -8 |
| ; AVX512VL-32-NEXT: movl %esp, %ebp |
| ; AVX512VL-32-NEXT: .cfi_def_cfa_register %ebp |
| ; AVX512VL-32-NEXT: andl $-8, %esp |
| ; AVX512VL-32-NEXT: subl $32, %esp |
| ; AVX512VL-32-NEXT: vmovlps %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: vmovhps %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: vextractf128 $1, %ymm0, %xmm0 |
| ; AVX512VL-32-NEXT: vmovlps %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: vmovhps %xmm0, (%esp) |
| ; AVX512VL-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fldl (%esp) |
| ; AVX512VL-32-NEXT: fisttpll (%esp) |
| ; AVX512VL-32-NEXT: wait |
| ; AVX512VL-32-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; AVX512VL-32-NEXT: vpinsrd $1, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX512VL-32-NEXT: vpinsrd $2, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX512VL-32-NEXT: vpinsrd $3, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX512VL-32-NEXT: vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX512VL-32-NEXT: vpinsrd $1, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX512VL-32-NEXT: vpinsrd $2, (%esp), %xmm1, %xmm1 |
| ; AVX512VL-32-NEXT: vpinsrd $3, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX512VL-32-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX512VL-32-NEXT: movl %ebp, %esp |
| ; AVX512VL-32-NEXT: popl %ebp |
| ; AVX512VL-32-NEXT: .cfi_def_cfa %esp, 4 |
| ; AVX512VL-32-NEXT: retl |
| ; |
| ; AVX512VL-64-LABEL: strict_vector_fptosi_v4f64_to_v4i64: |
| ; AVX512VL-64: # %bb.0: |
| ; AVX512VL-64-NEXT: vextractf128 $1, %ymm0, %xmm1 |
| ; AVX512VL-64-NEXT: vcvttsd2si %xmm1, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0] |
| ; AVX512VL-64-NEXT: vcvttsd2si %xmm1, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm1 |
| ; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0] |
| ; AVX512VL-64-NEXT: vcvttsd2si %xmm0, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] |
| ; AVX512VL-64-NEXT: vcvttsd2si %xmm0, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm0 |
| ; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0] |
| ; AVX512VL-64-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX512VL-64-NEXT: retq |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptosi_v4f64_to_v4i64: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vmovaps %ymm0, %ymm0 |
| ; AVX512DQ-NEXT: vcvttpd2qq %zmm0, %zmm0 |
| ; AVX512DQ-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0 |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptosi_v4f64_to_v4i64: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttpd2qq %ymm0, %ymm0 |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <4 x i64> @llvm.experimental.constrained.fptosi.v4i64.v4f64(<4 x double> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <4 x i64> %ret |
| } |
| |
| define <4 x i64> @strict_vector_fptoui_v4f64_to_v4i64(<4 x double> %a) #0 { |
| ; AVX-32-LABEL: strict_vector_fptoui_v4f64_to_v4i64: |
| ; AVX-32: # %bb.0: |
| ; AVX-32-NEXT: pushl %ebp |
| ; AVX-32-NEXT: .cfi_def_cfa_offset 8 |
| ; AVX-32-NEXT: .cfi_offset %ebp, -8 |
| ; AVX-32-NEXT: movl %esp, %ebp |
| ; AVX-32-NEXT: .cfi_def_cfa_register %ebp |
| ; AVX-32-NEXT: andl $-8, %esp |
| ; AVX-32-NEXT: subl $32, %esp |
| ; AVX-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0] |
| ; AVX-32-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero |
| ; AVX-32-NEXT: vcomisd %xmm1, %xmm2 |
| ; AVX-32-NEXT: vmovapd %xmm1, %xmm3 |
| ; AVX-32-NEXT: jae .LBB1_2 |
| ; AVX-32-NEXT: # %bb.1: |
| ; AVX-32-NEXT: vxorpd %xmm3, %xmm3, %xmm3 |
| ; AVX-32-NEXT: .LBB1_2: |
| ; AVX-32-NEXT: vsubsd %xmm3, %xmm2, %xmm2 |
| ; AVX-32-NEXT: vmovsd %xmm2, {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: wait |
| ; AVX-32-NEXT: setae %al |
| ; AVX-32-NEXT: movzbl %al, %eax |
| ; AVX-32-NEXT: shll $31, %eax |
| ; AVX-32-NEXT: xorl {{[0-9]+}}(%esp), %eax |
| ; AVX-32-NEXT: vextractf128 $1, %ymm0, %xmm2 |
| ; AVX-32-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0] |
| ; AVX-32-NEXT: vcomisd %xmm1, %xmm3 |
| ; AVX-32-NEXT: vmovapd %xmm1, %xmm4 |
| ; AVX-32-NEXT: jae .LBB1_4 |
| ; AVX-32-NEXT: # %bb.3: |
| ; AVX-32-NEXT: vxorpd %xmm4, %xmm4, %xmm4 |
| ; AVX-32-NEXT: .LBB1_4: |
| ; AVX-32-NEXT: vsubsd %xmm4, %xmm3, %xmm3 |
| ; AVX-32-NEXT: vmovsd %xmm3, (%esp) |
| ; AVX-32-NEXT: fldl (%esp) |
| ; AVX-32-NEXT: fisttpll (%esp) |
| ; AVX-32-NEXT: wait |
| ; AVX-32-NEXT: setae %cl |
| ; AVX-32-NEXT: movzbl %cl, %ecx |
| ; AVX-32-NEXT: shll $31, %ecx |
| ; AVX-32-NEXT: xorl {{[0-9]+}}(%esp), %ecx |
| ; AVX-32-NEXT: vcomisd %xmm1, %xmm2 |
| ; AVX-32-NEXT: vmovapd %xmm1, %xmm3 |
| ; AVX-32-NEXT: jae .LBB1_6 |
| ; AVX-32-NEXT: # %bb.5: |
| ; AVX-32-NEXT: vxorpd %xmm3, %xmm3, %xmm3 |
| ; AVX-32-NEXT: .LBB1_6: |
| ; AVX-32-NEXT: vsubsd %xmm3, %xmm2, %xmm2 |
| ; AVX-32-NEXT: vmovsd %xmm2, {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: wait |
| ; AVX-32-NEXT: setae %dl |
| ; AVX-32-NEXT: movzbl %dl, %edx |
| ; AVX-32-NEXT: shll $31, %edx |
| ; AVX-32-NEXT: xorl {{[0-9]+}}(%esp), %edx |
| ; AVX-32-NEXT: vcomisd %xmm1, %xmm0 |
| ; AVX-32-NEXT: jae .LBB1_8 |
| ; AVX-32-NEXT: # %bb.7: |
| ; AVX-32-NEXT: vxorpd %xmm1, %xmm1, %xmm1 |
| ; AVX-32-NEXT: .LBB1_8: |
| ; AVX-32-NEXT: vsubsd %xmm1, %xmm0, %xmm0 |
| ; AVX-32-NEXT: vmovsd %xmm0, {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: wait |
| ; AVX-32-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; AVX-32-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 |
| ; AVX-32-NEXT: vpinsrd $2, (%esp), %xmm0, %xmm0 |
| ; AVX-32-NEXT: vpinsrd $3, %ecx, %xmm0, %xmm0 |
| ; AVX-32-NEXT: setae %cl |
| ; AVX-32-NEXT: movzbl %cl, %ecx |
| ; AVX-32-NEXT: shll $31, %ecx |
| ; AVX-32-NEXT: xorl {{[0-9]+}}(%esp), %ecx |
| ; AVX-32-NEXT: vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX-32-NEXT: vpinsrd $1, %ecx, %xmm1, %xmm1 |
| ; AVX-32-NEXT: vpinsrd $2, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX-32-NEXT: vpinsrd $3, %eax, %xmm1, %xmm1 |
| ; AVX-32-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0 |
| ; AVX-32-NEXT: movl %ebp, %esp |
| ; AVX-32-NEXT: popl %ebp |
| ; AVX-32-NEXT: .cfi_def_cfa %esp, 4 |
| ; AVX-32-NEXT: retl |
| ; |
| ; AVX-64-LABEL: strict_vector_fptoui_v4f64_to_v4i64: |
| ; AVX-64: # %bb.0: |
| ; AVX-64-NEXT: vextractf128 $1, %ymm0, %xmm3 |
| ; AVX-64-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero |
| ; AVX-64-NEXT: vcomisd %xmm1, %xmm3 |
| ; AVX-64-NEXT: vxorpd %xmm2, %xmm2, %xmm2 |
| ; AVX-64-NEXT: vxorpd %xmm4, %xmm4, %xmm4 |
| ; AVX-64-NEXT: jb .LBB1_2 |
| ; AVX-64-NEXT: # %bb.1: |
| ; AVX-64-NEXT: vmovapd %xmm1, %xmm4 |
| ; AVX-64-NEXT: .LBB1_2: |
| ; AVX-64-NEXT: vsubsd %xmm4, %xmm3, %xmm4 |
| ; AVX-64-NEXT: vcvttsd2si %xmm4, %rcx |
| ; AVX-64-NEXT: setae %al |
| ; AVX-64-NEXT: movzbl %al, %eax |
| ; AVX-64-NEXT: shlq $63, %rax |
| ; AVX-64-NEXT: xorq %rcx, %rax |
| ; AVX-64-NEXT: vpermilpd {{.*#+}} xmm4 = xmm3[1,0] |
| ; AVX-64-NEXT: vcomisd %xmm1, %xmm4 |
| ; AVX-64-NEXT: vxorpd %xmm5, %xmm5, %xmm5 |
| ; AVX-64-NEXT: jb .LBB1_4 |
| ; AVX-64-NEXT: # %bb.3: |
| ; AVX-64-NEXT: vmovapd %xmm1, %xmm5 |
| ; AVX-64-NEXT: .LBB1_4: |
| ; AVX-64-NEXT: vmovq %rax, %xmm3 |
| ; AVX-64-NEXT: vsubsd %xmm5, %xmm4, %xmm4 |
| ; AVX-64-NEXT: vcvttsd2si %xmm4, %rax |
| ; AVX-64-NEXT: setae %cl |
| ; AVX-64-NEXT: movzbl %cl, %ecx |
| ; AVX-64-NEXT: shlq $63, %rcx |
| ; AVX-64-NEXT: xorq %rax, %rcx |
| ; AVX-64-NEXT: vmovq %rcx, %xmm4 |
| ; AVX-64-NEXT: vcomisd %xmm1, %xmm0 |
| ; AVX-64-NEXT: vxorpd %xmm5, %xmm5, %xmm5 |
| ; AVX-64-NEXT: jb .LBB1_6 |
| ; AVX-64-NEXT: # %bb.5: |
| ; AVX-64-NEXT: vmovapd %xmm1, %xmm5 |
| ; AVX-64-NEXT: .LBB1_6: |
| ; AVX-64-NEXT: vpunpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm4[0] |
| ; AVX-64-NEXT: vsubsd %xmm5, %xmm0, %xmm4 |
| ; AVX-64-NEXT: vcvttsd2si %xmm4, %rax |
| ; AVX-64-NEXT: setae %cl |
| ; AVX-64-NEXT: movzbl %cl, %ecx |
| ; AVX-64-NEXT: shlq $63, %rcx |
| ; AVX-64-NEXT: xorq %rax, %rcx |
| ; AVX-64-NEXT: vmovq %rcx, %xmm4 |
| ; AVX-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] |
| ; AVX-64-NEXT: vcomisd %xmm1, %xmm0 |
| ; AVX-64-NEXT: jb .LBB1_8 |
| ; AVX-64-NEXT: # %bb.7: |
| ; AVX-64-NEXT: vmovapd %xmm1, %xmm2 |
| ; AVX-64-NEXT: .LBB1_8: |
| ; AVX-64-NEXT: vsubsd %xmm2, %xmm0, %xmm0 |
| ; AVX-64-NEXT: vcvttsd2si %xmm0, %rax |
| ; AVX-64-NEXT: setae %cl |
| ; AVX-64-NEXT: movzbl %cl, %ecx |
| ; AVX-64-NEXT: shlq $63, %rcx |
| ; AVX-64-NEXT: xorq %rax, %rcx |
| ; AVX-64-NEXT: vmovq %rcx, %xmm0 |
| ; AVX-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm4[0],xmm0[0] |
| ; AVX-64-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0 |
| ; AVX-64-NEXT: retq |
| ; |
| ; AVX512F-32-LABEL: strict_vector_fptoui_v4f64_to_v4i64: |
| ; AVX512F-32: # %bb.0: |
| ; AVX512F-32-NEXT: pushl %ebp |
| ; AVX512F-32-NEXT: .cfi_def_cfa_offset 8 |
| ; AVX512F-32-NEXT: .cfi_offset %ebp, -8 |
| ; AVX512F-32-NEXT: movl %esp, %ebp |
| ; AVX512F-32-NEXT: .cfi_def_cfa_register %ebp |
| ; AVX512F-32-NEXT: pushl %ebx |
| ; AVX512F-32-NEXT: andl $-8, %esp |
| ; AVX512F-32-NEXT: subl $40, %esp |
| ; AVX512F-32-NEXT: .cfi_offset %ebx, -12 |
| ; AVX512F-32-NEXT: vextractf128 $1, %ymm0, %xmm2 |
| ; AVX512F-32-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0] |
| ; AVX512F-32-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero |
| ; AVX512F-32-NEXT: xorl %eax, %eax |
| ; AVX512F-32-NEXT: vcomisd %xmm1, %xmm3 |
| ; AVX512F-32-NEXT: setae %al |
| ; AVX512F-32-NEXT: kmovw %eax, %k1 |
| ; AVX512F-32-NEXT: vmovsd %xmm1, %xmm1, %xmm4 {%k1} {z} |
| ; AVX512F-32-NEXT: vsubsd %xmm4, %xmm3, %xmm3 |
| ; AVX512F-32-NEXT: vmovsd %xmm3, (%esp) |
| ; AVX512F-32-NEXT: xorl %edx, %edx |
| ; AVX512F-32-NEXT: vcomisd %xmm1, %xmm2 |
| ; AVX512F-32-NEXT: setae %dl |
| ; AVX512F-32-NEXT: kmovw %edx, %k1 |
| ; AVX512F-32-NEXT: vmovsd %xmm1, %xmm1, %xmm3 {%k1} {z} |
| ; AVX512F-32-NEXT: vsubsd %xmm3, %xmm2, %xmm2 |
| ; AVX512F-32-NEXT: vmovsd %xmm2, {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0] |
| ; AVX512F-32-NEXT: xorl %ecx, %ecx |
| ; AVX512F-32-NEXT: vcomisd %xmm1, %xmm2 |
| ; AVX512F-32-NEXT: setae %cl |
| ; AVX512F-32-NEXT: kmovw %ecx, %k1 |
| ; AVX512F-32-NEXT: vmovsd %xmm1, %xmm1, %xmm3 {%k1} {z} |
| ; AVX512F-32-NEXT: vsubsd %xmm3, %xmm2, %xmm2 |
| ; AVX512F-32-NEXT: vmovsd %xmm2, {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: xorl %ebx, %ebx |
| ; AVX512F-32-NEXT: vcomisd %xmm1, %xmm0 |
| ; AVX512F-32-NEXT: setae %bl |
| ; AVX512F-32-NEXT: kmovw %ebx, %k1 |
| ; AVX512F-32-NEXT: vmovsd %xmm1, %xmm1, %xmm1 {%k1} {z} |
| ; AVX512F-32-NEXT: vsubsd %xmm1, %xmm0, %xmm0 |
| ; AVX512F-32-NEXT: vmovsd %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fldl (%esp) |
| ; AVX512F-32-NEXT: fisttpll (%esp) |
| ; AVX512F-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: wait |
| ; AVX512F-32-NEXT: shll $31, %eax |
| ; AVX512F-32-NEXT: xorl {{[0-9]+}}(%esp), %eax |
| ; AVX512F-32-NEXT: shll $31, %edx |
| ; AVX512F-32-NEXT: xorl {{[0-9]+}}(%esp), %edx |
| ; AVX512F-32-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; AVX512F-32-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 |
| ; AVX512F-32-NEXT: vpinsrd $2, (%esp), %xmm0, %xmm0 |
| ; AVX512F-32-NEXT: vpinsrd $3, %eax, %xmm0, %xmm0 |
| ; AVX512F-32-NEXT: shll $31, %ecx |
| ; AVX512F-32-NEXT: xorl {{[0-9]+}}(%esp), %ecx |
| ; AVX512F-32-NEXT: shll $31, %ebx |
| ; AVX512F-32-NEXT: xorl {{[0-9]+}}(%esp), %ebx |
| ; AVX512F-32-NEXT: vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX512F-32-NEXT: vpinsrd $1, %ebx, %xmm1, %xmm1 |
| ; AVX512F-32-NEXT: vpinsrd $2, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX512F-32-NEXT: vpinsrd $3, %ecx, %xmm1, %xmm1 |
| ; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 |
| ; AVX512F-32-NEXT: leal -4(%ebp), %esp |
| ; AVX512F-32-NEXT: popl %ebx |
| ; AVX512F-32-NEXT: popl %ebp |
| ; AVX512F-32-NEXT: .cfi_def_cfa %esp, 4 |
| ; AVX512F-32-NEXT: retl |
| ; |
| ; AVX512F-64-LABEL: strict_vector_fptoui_v4f64_to_v4i64: |
| ; AVX512F-64: # %bb.0: |
| ; AVX512F-64-NEXT: vextractf128 $1, %ymm0, %xmm1 |
| ; AVX512F-64-NEXT: vcvttsd2usi %xmm1, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512F-64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0] |
| ; AVX512F-64-NEXT: vcvttsd2usi %xmm1, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm1 |
| ; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0] |
| ; AVX512F-64-NEXT: vcvttsd2usi %xmm0, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512F-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] |
| ; AVX512F-64-NEXT: vcvttsd2usi %xmm0, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm0 |
| ; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0] |
| ; AVX512F-64-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX512F-64-NEXT: retq |
| ; |
| ; AVX512VL-32-LABEL: strict_vector_fptoui_v4f64_to_v4i64: |
| ; AVX512VL-32: # %bb.0: |
| ; AVX512VL-32-NEXT: pushl %ebp |
| ; AVX512VL-32-NEXT: .cfi_def_cfa_offset 8 |
| ; AVX512VL-32-NEXT: .cfi_offset %ebp, -8 |
| ; AVX512VL-32-NEXT: movl %esp, %ebp |
| ; AVX512VL-32-NEXT: .cfi_def_cfa_register %ebp |
| ; AVX512VL-32-NEXT: pushl %ebx |
| ; AVX512VL-32-NEXT: andl $-8, %esp |
| ; AVX512VL-32-NEXT: subl $40, %esp |
| ; AVX512VL-32-NEXT: .cfi_offset %ebx, -12 |
| ; AVX512VL-32-NEXT: vextractf128 $1, %ymm0, %xmm2 |
| ; AVX512VL-32-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0] |
| ; AVX512VL-32-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero |
| ; AVX512VL-32-NEXT: xorl %eax, %eax |
| ; AVX512VL-32-NEXT: vcomisd %xmm1, %xmm3 |
| ; AVX512VL-32-NEXT: setae %al |
| ; AVX512VL-32-NEXT: kmovw %eax, %k1 |
| ; AVX512VL-32-NEXT: vmovsd %xmm1, %xmm1, %xmm4 {%k1} {z} |
| ; AVX512VL-32-NEXT: vsubsd %xmm4, %xmm3, %xmm3 |
| ; AVX512VL-32-NEXT: vmovsd %xmm3, (%esp) |
| ; AVX512VL-32-NEXT: xorl %edx, %edx |
| ; AVX512VL-32-NEXT: vcomisd %xmm1, %xmm2 |
| ; AVX512VL-32-NEXT: setae %dl |
| ; AVX512VL-32-NEXT: kmovw %edx, %k1 |
| ; AVX512VL-32-NEXT: vmovsd %xmm1, %xmm1, %xmm3 {%k1} {z} |
| ; AVX512VL-32-NEXT: vsubsd %xmm3, %xmm2, %xmm2 |
| ; AVX512VL-32-NEXT: vmovsd %xmm2, {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0] |
| ; AVX512VL-32-NEXT: xorl %ecx, %ecx |
| ; AVX512VL-32-NEXT: vcomisd %xmm1, %xmm2 |
| ; AVX512VL-32-NEXT: setae %cl |
| ; AVX512VL-32-NEXT: kmovw %ecx, %k1 |
| ; AVX512VL-32-NEXT: vmovsd %xmm1, %xmm1, %xmm3 {%k1} {z} |
| ; AVX512VL-32-NEXT: vsubsd %xmm3, %xmm2, %xmm2 |
| ; AVX512VL-32-NEXT: vmovsd %xmm2, {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: xorl %ebx, %ebx |
| ; AVX512VL-32-NEXT: vcomisd %xmm1, %xmm0 |
| ; AVX512VL-32-NEXT: setae %bl |
| ; AVX512VL-32-NEXT: kmovw %ebx, %k1 |
| ; AVX512VL-32-NEXT: vmovsd %xmm1, %xmm1, %xmm1 {%k1} {z} |
| ; AVX512VL-32-NEXT: vsubsd %xmm1, %xmm0, %xmm0 |
| ; AVX512VL-32-NEXT: vmovsd %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fldl (%esp) |
| ; AVX512VL-32-NEXT: fisttpll (%esp) |
| ; AVX512VL-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fldl {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: wait |
| ; AVX512VL-32-NEXT: shll $31, %eax |
| ; AVX512VL-32-NEXT: xorl {{[0-9]+}}(%esp), %eax |
| ; AVX512VL-32-NEXT: shll $31, %edx |
| ; AVX512VL-32-NEXT: xorl {{[0-9]+}}(%esp), %edx |
| ; AVX512VL-32-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; AVX512VL-32-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 |
| ; AVX512VL-32-NEXT: vpinsrd $2, (%esp), %xmm0, %xmm0 |
| ; AVX512VL-32-NEXT: vpinsrd $3, %eax, %xmm0, %xmm0 |
| ; AVX512VL-32-NEXT: shll $31, %ecx |
| ; AVX512VL-32-NEXT: xorl {{[0-9]+}}(%esp), %ecx |
| ; AVX512VL-32-NEXT: shll $31, %ebx |
| ; AVX512VL-32-NEXT: xorl {{[0-9]+}}(%esp), %ebx |
| ; AVX512VL-32-NEXT: vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX512VL-32-NEXT: vpinsrd $1, %ebx, %xmm1, %xmm1 |
| ; AVX512VL-32-NEXT: vpinsrd $2, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX512VL-32-NEXT: vpinsrd $3, %ecx, %xmm1, %xmm1 |
| ; AVX512VL-32-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 |
| ; AVX512VL-32-NEXT: leal -4(%ebp), %esp |
| ; AVX512VL-32-NEXT: popl %ebx |
| ; AVX512VL-32-NEXT: popl %ebp |
| ; AVX512VL-32-NEXT: .cfi_def_cfa %esp, 4 |
| ; AVX512VL-32-NEXT: retl |
| ; |
| ; AVX512VL-64-LABEL: strict_vector_fptoui_v4f64_to_v4i64: |
| ; AVX512VL-64: # %bb.0: |
| ; AVX512VL-64-NEXT: vextractf128 $1, %ymm0, %xmm1 |
| ; AVX512VL-64-NEXT: vcvttsd2usi %xmm1, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0] |
| ; AVX512VL-64-NEXT: vcvttsd2usi %xmm1, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm1 |
| ; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0] |
| ; AVX512VL-64-NEXT: vcvttsd2usi %xmm0, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0] |
| ; AVX512VL-64-NEXT: vcvttsd2usi %xmm0, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm0 |
| ; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0] |
| ; AVX512VL-64-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX512VL-64-NEXT: retq |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptoui_v4f64_to_v4i64: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vmovaps %ymm0, %ymm0 |
| ; AVX512DQ-NEXT: vcvttpd2uqq %zmm0, %zmm0 |
| ; AVX512DQ-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0 |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptoui_v4f64_to_v4i64: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttpd2uqq %ymm0, %ymm0 |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <4 x i64> @llvm.experimental.constrained.fptoui.v4i64.v4f64(<4 x double> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <4 x i64> %ret |
| } |
| |
| define <4 x i64> @strict_vector_fptosi_v4f32_to_v4i64(<4 x float> %a) #0 { |
| ; AVX-32-LABEL: strict_vector_fptosi_v4f32_to_v4i64: |
| ; AVX-32: # %bb.0: |
| ; AVX-32-NEXT: pushl %ebp |
| ; AVX-32-NEXT: .cfi_def_cfa_offset 8 |
| ; AVX-32-NEXT: .cfi_offset %ebp, -8 |
| ; AVX-32-NEXT: movl %esp, %ebp |
| ; AVX-32-NEXT: .cfi_def_cfa_register %ebp |
| ; AVX-32-NEXT: andl $-8, %esp |
| ; AVX-32-NEXT: subl $32, %esp |
| ; AVX-32-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: vextractps $2, %xmm0, {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: vextractps $3, %xmm0, (%esp) |
| ; AVX-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: flds (%esp) |
| ; AVX-32-NEXT: fisttpll (%esp) |
| ; AVX-32-NEXT: wait |
| ; AVX-32-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; AVX-32-NEXT: vpinsrd $1, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX-32-NEXT: vpinsrd $2, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX-32-NEXT: vpinsrd $3, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX-32-NEXT: vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX-32-NEXT: vpinsrd $1, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX-32-NEXT: vpinsrd $2, (%esp), %xmm1, %xmm1 |
| ; AVX-32-NEXT: vpinsrd $3, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX-32-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX-32-NEXT: movl %ebp, %esp |
| ; AVX-32-NEXT: popl %ebp |
| ; AVX-32-NEXT: .cfi_def_cfa %esp, 4 |
| ; AVX-32-NEXT: retl |
| ; |
| ; AVX-64-LABEL: strict_vector_fptosi_v4f32_to_v4i64: |
| ; AVX-64: # %bb.0: |
| ; AVX-64-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3] |
| ; AVX-64-NEXT: vcvttss2si %xmm1, %rax |
| ; AVX-64-NEXT: vmovq %rax, %xmm1 |
| ; AVX-64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0] |
| ; AVX-64-NEXT: vcvttss2si %xmm2, %rax |
| ; AVX-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0] |
| ; AVX-64-NEXT: vcvttss2si %xmm0, %rax |
| ; AVX-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX-64-NEXT: vmovshdup {{.*#+}} xmm0 = xmm0[1,1,3,3] |
| ; AVX-64-NEXT: vcvttss2si %xmm0, %rax |
| ; AVX-64-NEXT: vmovq %rax, %xmm0 |
| ; AVX-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0] |
| ; AVX-64-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX-64-NEXT: retq |
| ; |
| ; AVX512F-32-LABEL: strict_vector_fptosi_v4f32_to_v4i64: |
| ; AVX512F-32: # %bb.0: |
| ; AVX512F-32-NEXT: pushl %ebp |
| ; AVX512F-32-NEXT: .cfi_def_cfa_offset 8 |
| ; AVX512F-32-NEXT: .cfi_offset %ebp, -8 |
| ; AVX512F-32-NEXT: movl %esp, %ebp |
| ; AVX512F-32-NEXT: .cfi_def_cfa_register %ebp |
| ; AVX512F-32-NEXT: andl $-8, %esp |
| ; AVX512F-32-NEXT: subl $32, %esp |
| ; AVX512F-32-NEXT: vmovd %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: vextractps $2, %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: vextractps $3, %xmm0, (%esp) |
| ; AVX512F-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: flds (%esp) |
| ; AVX512F-32-NEXT: fisttpll (%esp) |
| ; AVX512F-32-NEXT: wait |
| ; AVX512F-32-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; AVX512F-32-NEXT: vpinsrd $1, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX512F-32-NEXT: vpinsrd $2, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX512F-32-NEXT: vpinsrd $3, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX512F-32-NEXT: vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX512F-32-NEXT: vpinsrd $1, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX512F-32-NEXT: vpinsrd $2, (%esp), %xmm1, %xmm1 |
| ; AVX512F-32-NEXT: vpinsrd $3, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX512F-32-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX512F-32-NEXT: movl %ebp, %esp |
| ; AVX512F-32-NEXT: popl %ebp |
| ; AVX512F-32-NEXT: .cfi_def_cfa %esp, 4 |
| ; AVX512F-32-NEXT: retl |
| ; |
| ; AVX512F-64-LABEL: strict_vector_fptosi_v4f32_to_v4i64: |
| ; AVX512F-64: # %bb.0: |
| ; AVX512F-64-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3] |
| ; AVX512F-64-NEXT: vcvttss2si %xmm1, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm1 |
| ; AVX512F-64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0] |
| ; AVX512F-64-NEXT: vcvttss2si %xmm2, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0] |
| ; AVX512F-64-NEXT: vcvttss2si %xmm0, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512F-64-NEXT: vmovshdup {{.*#+}} xmm0 = xmm0[1,1,3,3] |
| ; AVX512F-64-NEXT: vcvttss2si %xmm0, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm0 |
| ; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0] |
| ; AVX512F-64-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX512F-64-NEXT: retq |
| ; |
| ; AVX512VL-32-LABEL: strict_vector_fptosi_v4f32_to_v4i64: |
| ; AVX512VL-32: # %bb.0: |
| ; AVX512VL-32-NEXT: pushl %ebp |
| ; AVX512VL-32-NEXT: .cfi_def_cfa_offset 8 |
| ; AVX512VL-32-NEXT: .cfi_offset %ebp, -8 |
| ; AVX512VL-32-NEXT: movl %esp, %ebp |
| ; AVX512VL-32-NEXT: .cfi_def_cfa_register %ebp |
| ; AVX512VL-32-NEXT: andl $-8, %esp |
| ; AVX512VL-32-NEXT: subl $32, %esp |
| ; AVX512VL-32-NEXT: vmovd %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: vextractps $1, %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: vextractps $2, %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: vextractps $3, %xmm0, (%esp) |
| ; AVX512VL-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: flds (%esp) |
| ; AVX512VL-32-NEXT: fisttpll (%esp) |
| ; AVX512VL-32-NEXT: wait |
| ; AVX512VL-32-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; AVX512VL-32-NEXT: vpinsrd $1, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX512VL-32-NEXT: vpinsrd $2, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX512VL-32-NEXT: vpinsrd $3, {{[0-9]+}}(%esp), %xmm0, %xmm0 |
| ; AVX512VL-32-NEXT: vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX512VL-32-NEXT: vpinsrd $1, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX512VL-32-NEXT: vpinsrd $2, (%esp), %xmm1, %xmm1 |
| ; AVX512VL-32-NEXT: vpinsrd $3, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX512VL-32-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX512VL-32-NEXT: movl %ebp, %esp |
| ; AVX512VL-32-NEXT: popl %ebp |
| ; AVX512VL-32-NEXT: .cfi_def_cfa %esp, 4 |
| ; AVX512VL-32-NEXT: retl |
| ; |
| ; AVX512VL-64-LABEL: strict_vector_fptosi_v4f32_to_v4i64: |
| ; AVX512VL-64: # %bb.0: |
| ; AVX512VL-64-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3] |
| ; AVX512VL-64-NEXT: vcvttss2si %xmm1, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm1 |
| ; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0] |
| ; AVX512VL-64-NEXT: vcvttss2si %xmm2, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0] |
| ; AVX512VL-64-NEXT: vcvttss2si %xmm0, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512VL-64-NEXT: vmovshdup {{.*#+}} xmm0 = xmm0[1,1,3,3] |
| ; AVX512VL-64-NEXT: vcvttss2si %xmm0, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm0 |
| ; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0] |
| ; AVX512VL-64-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX512VL-64-NEXT: retq |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptosi_v4f32_to_v4i64: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vmovaps %xmm0, %xmm0 |
| ; AVX512DQ-NEXT: vcvttps2qq %ymm0, %zmm0 |
| ; AVX512DQ-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0 |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptosi_v4f32_to_v4i64: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttps2qq %xmm0, %ymm0 |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <4 x i64> @llvm.experimental.constrained.fptosi.v4i64.v4f32(<4 x float> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <4 x i64> %ret |
| } |
| |
| define <4 x i64> @strict_vector_fptoui_v4f32_to_v4i64(<4 x float> %a) #0 { |
| ; AVX-32-LABEL: strict_vector_fptoui_v4f32_to_v4i64: |
| ; AVX-32: # %bb.0: |
| ; AVX-32-NEXT: pushl %ebp |
| ; AVX-32-NEXT: .cfi_def_cfa_offset 8 |
| ; AVX-32-NEXT: .cfi_offset %ebp, -8 |
| ; AVX-32-NEXT: movl %esp, %ebp |
| ; AVX-32-NEXT: .cfi_def_cfa_register %ebp |
| ; AVX-32-NEXT: andl $-8, %esp |
| ; AVX-32-NEXT: subl $32, %esp |
| ; AVX-32-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3] |
| ; AVX-32-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX-32-NEXT: vcomiss %xmm1, %xmm2 |
| ; AVX-32-NEXT: vmovaps %xmm1, %xmm3 |
| ; AVX-32-NEXT: jae .LBB3_2 |
| ; AVX-32-NEXT: # %bb.1: |
| ; AVX-32-NEXT: vxorps %xmm3, %xmm3, %xmm3 |
| ; AVX-32-NEXT: .LBB3_2: |
| ; AVX-32-NEXT: vsubss %xmm3, %xmm2, %xmm2 |
| ; AVX-32-NEXT: vmovss %xmm2, {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: wait |
| ; AVX-32-NEXT: setae %al |
| ; AVX-32-NEXT: movzbl %al, %eax |
| ; AVX-32-NEXT: shll $31, %eax |
| ; AVX-32-NEXT: xorl {{[0-9]+}}(%esp), %eax |
| ; AVX-32-NEXT: vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3] |
| ; AVX-32-NEXT: vcomiss %xmm1, %xmm2 |
| ; AVX-32-NEXT: vmovaps %xmm1, %xmm3 |
| ; AVX-32-NEXT: jae .LBB3_4 |
| ; AVX-32-NEXT: # %bb.3: |
| ; AVX-32-NEXT: vxorps %xmm3, %xmm3, %xmm3 |
| ; AVX-32-NEXT: .LBB3_4: |
| ; AVX-32-NEXT: vsubss %xmm3, %xmm2, %xmm2 |
| ; AVX-32-NEXT: vmovss %xmm2, (%esp) |
| ; AVX-32-NEXT: flds (%esp) |
| ; AVX-32-NEXT: fisttpll (%esp) |
| ; AVX-32-NEXT: wait |
| ; AVX-32-NEXT: setae %cl |
| ; AVX-32-NEXT: movzbl %cl, %ecx |
| ; AVX-32-NEXT: shll $31, %ecx |
| ; AVX-32-NEXT: xorl {{[0-9]+}}(%esp), %ecx |
| ; AVX-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0] |
| ; AVX-32-NEXT: vcomiss %xmm1, %xmm2 |
| ; AVX-32-NEXT: vmovaps %xmm1, %xmm3 |
| ; AVX-32-NEXT: jae .LBB3_6 |
| ; AVX-32-NEXT: # %bb.5: |
| ; AVX-32-NEXT: vxorps %xmm3, %xmm3, %xmm3 |
| ; AVX-32-NEXT: .LBB3_6: |
| ; AVX-32-NEXT: vsubss %xmm3, %xmm2, %xmm2 |
| ; AVX-32-NEXT: vmovss %xmm2, {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: wait |
| ; AVX-32-NEXT: setae %dl |
| ; AVX-32-NEXT: movzbl %dl, %edx |
| ; AVX-32-NEXT: shll $31, %edx |
| ; AVX-32-NEXT: xorl {{[0-9]+}}(%esp), %edx |
| ; AVX-32-NEXT: vcomiss %xmm1, %xmm0 |
| ; AVX-32-NEXT: jae .LBB3_8 |
| ; AVX-32-NEXT: # %bb.7: |
| ; AVX-32-NEXT: vxorps %xmm1, %xmm1, %xmm1 |
| ; AVX-32-NEXT: .LBB3_8: |
| ; AVX-32-NEXT: vsubss %xmm1, %xmm0, %xmm0 |
| ; AVX-32-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX-32-NEXT: wait |
| ; AVX-32-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; AVX-32-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 |
| ; AVX-32-NEXT: vpinsrd $2, (%esp), %xmm0, %xmm0 |
| ; AVX-32-NEXT: vpinsrd $3, %ecx, %xmm0, %xmm0 |
| ; AVX-32-NEXT: setae %cl |
| ; AVX-32-NEXT: movzbl %cl, %ecx |
| ; AVX-32-NEXT: shll $31, %ecx |
| ; AVX-32-NEXT: xorl {{[0-9]+}}(%esp), %ecx |
| ; AVX-32-NEXT: vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX-32-NEXT: vpinsrd $1, %ecx, %xmm1, %xmm1 |
| ; AVX-32-NEXT: vpinsrd $2, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX-32-NEXT: vpinsrd $3, %eax, %xmm1, %xmm1 |
| ; AVX-32-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0 |
| ; AVX-32-NEXT: movl %ebp, %esp |
| ; AVX-32-NEXT: popl %ebp |
| ; AVX-32-NEXT: .cfi_def_cfa %esp, 4 |
| ; AVX-32-NEXT: retl |
| ; |
| ; AVX-64-LABEL: strict_vector_fptoui_v4f32_to_v4i64: |
| ; AVX-64: # %bb.0: |
| ; AVX-64-NEXT: vpermilps {{.*#+}} xmm3 = xmm0[3,3,3,3] |
| ; AVX-64-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX-64-NEXT: vcomiss %xmm1, %xmm3 |
| ; AVX-64-NEXT: vxorps %xmm2, %xmm2, %xmm2 |
| ; AVX-64-NEXT: vxorps %xmm4, %xmm4, %xmm4 |
| ; AVX-64-NEXT: jb .LBB3_2 |
| ; AVX-64-NEXT: # %bb.1: |
| ; AVX-64-NEXT: vmovaps %xmm1, %xmm4 |
| ; AVX-64-NEXT: .LBB3_2: |
| ; AVX-64-NEXT: vsubss %xmm4, %xmm3, %xmm3 |
| ; AVX-64-NEXT: vcvttss2si %xmm3, %rcx |
| ; AVX-64-NEXT: setae %al |
| ; AVX-64-NEXT: movzbl %al, %eax |
| ; AVX-64-NEXT: shlq $63, %rax |
| ; AVX-64-NEXT: xorq %rcx, %rax |
| ; AVX-64-NEXT: vpermilpd {{.*#+}} xmm4 = xmm0[1,0] |
| ; AVX-64-NEXT: vcomiss %xmm1, %xmm4 |
| ; AVX-64-NEXT: vxorps %xmm5, %xmm5, %xmm5 |
| ; AVX-64-NEXT: jb .LBB3_4 |
| ; AVX-64-NEXT: # %bb.3: |
| ; AVX-64-NEXT: vmovaps %xmm1, %xmm5 |
| ; AVX-64-NEXT: .LBB3_4: |
| ; AVX-64-NEXT: vmovq %rax, %xmm3 |
| ; AVX-64-NEXT: vsubss %xmm5, %xmm4, %xmm4 |
| ; AVX-64-NEXT: vcvttss2si %xmm4, %rax |
| ; AVX-64-NEXT: setae %cl |
| ; AVX-64-NEXT: movzbl %cl, %ecx |
| ; AVX-64-NEXT: shlq $63, %rcx |
| ; AVX-64-NEXT: xorq %rax, %rcx |
| ; AVX-64-NEXT: vmovq %rcx, %xmm4 |
| ; AVX-64-NEXT: vcomiss %xmm1, %xmm0 |
| ; AVX-64-NEXT: vxorps %xmm5, %xmm5, %xmm5 |
| ; AVX-64-NEXT: jb .LBB3_6 |
| ; AVX-64-NEXT: # %bb.5: |
| ; AVX-64-NEXT: vmovaps %xmm1, %xmm5 |
| ; AVX-64-NEXT: .LBB3_6: |
| ; AVX-64-NEXT: vpunpcklqdq {{.*#+}} xmm3 = xmm4[0],xmm3[0] |
| ; AVX-64-NEXT: vsubss %xmm5, %xmm0, %xmm4 |
| ; AVX-64-NEXT: vcvttss2si %xmm4, %rax |
| ; AVX-64-NEXT: setae %cl |
| ; AVX-64-NEXT: movzbl %cl, %ecx |
| ; AVX-64-NEXT: shlq $63, %rcx |
| ; AVX-64-NEXT: xorq %rax, %rcx |
| ; AVX-64-NEXT: vmovq %rcx, %xmm4 |
| ; AVX-64-NEXT: vmovshdup {{.*#+}} xmm0 = xmm0[1,1,3,3] |
| ; AVX-64-NEXT: vcomiss %xmm1, %xmm0 |
| ; AVX-64-NEXT: jb .LBB3_8 |
| ; AVX-64-NEXT: # %bb.7: |
| ; AVX-64-NEXT: vmovaps %xmm1, %xmm2 |
| ; AVX-64-NEXT: .LBB3_8: |
| ; AVX-64-NEXT: vsubss %xmm2, %xmm0, %xmm0 |
| ; AVX-64-NEXT: vcvttss2si %xmm0, %rax |
| ; AVX-64-NEXT: setae %cl |
| ; AVX-64-NEXT: movzbl %cl, %ecx |
| ; AVX-64-NEXT: shlq $63, %rcx |
| ; AVX-64-NEXT: xorq %rax, %rcx |
| ; AVX-64-NEXT: vmovq %rcx, %xmm0 |
| ; AVX-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm4[0],xmm0[0] |
| ; AVX-64-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0 |
| ; AVX-64-NEXT: retq |
| ; |
| ; AVX512F-32-LABEL: strict_vector_fptoui_v4f32_to_v4i64: |
| ; AVX512F-32: # %bb.0: |
| ; AVX512F-32-NEXT: pushl %ebp |
| ; AVX512F-32-NEXT: .cfi_def_cfa_offset 8 |
| ; AVX512F-32-NEXT: .cfi_offset %ebp, -8 |
| ; AVX512F-32-NEXT: movl %esp, %ebp |
| ; AVX512F-32-NEXT: .cfi_def_cfa_register %ebp |
| ; AVX512F-32-NEXT: pushl %ebx |
| ; AVX512F-32-NEXT: andl $-8, %esp |
| ; AVX512F-32-NEXT: subl $40, %esp |
| ; AVX512F-32-NEXT: .cfi_offset %ebx, -12 |
| ; AVX512F-32-NEXT: vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3] |
| ; AVX512F-32-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX512F-32-NEXT: xorl %eax, %eax |
| ; AVX512F-32-NEXT: vcomiss %xmm1, %xmm2 |
| ; AVX512F-32-NEXT: setae %al |
| ; AVX512F-32-NEXT: kmovw %eax, %k1 |
| ; AVX512F-32-NEXT: vmovss %xmm1, %xmm1, %xmm3 {%k1} {z} |
| ; AVX512F-32-NEXT: vsubss %xmm3, %xmm2, %xmm2 |
| ; AVX512F-32-NEXT: vmovss %xmm2, (%esp) |
| ; AVX512F-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0] |
| ; AVX512F-32-NEXT: xorl %edx, %edx |
| ; AVX512F-32-NEXT: vcomiss %xmm1, %xmm2 |
| ; AVX512F-32-NEXT: setae %dl |
| ; AVX512F-32-NEXT: kmovw %edx, %k1 |
| ; AVX512F-32-NEXT: vmovss %xmm1, %xmm1, %xmm3 {%k1} {z} |
| ; AVX512F-32-NEXT: vsubss %xmm3, %xmm2, %xmm2 |
| ; AVX512F-32-NEXT: vmovss %xmm2, {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3] |
| ; AVX512F-32-NEXT: xorl %ecx, %ecx |
| ; AVX512F-32-NEXT: vcomiss %xmm1, %xmm2 |
| ; AVX512F-32-NEXT: setae %cl |
| ; AVX512F-32-NEXT: kmovw %ecx, %k1 |
| ; AVX512F-32-NEXT: vmovss %xmm1, %xmm1, %xmm3 {%k1} {z} |
| ; AVX512F-32-NEXT: vsubss %xmm3, %xmm2, %xmm2 |
| ; AVX512F-32-NEXT: vmovss %xmm2, {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: xorl %ebx, %ebx |
| ; AVX512F-32-NEXT: vcomiss %xmm1, %xmm0 |
| ; AVX512F-32-NEXT: setae %bl |
| ; AVX512F-32-NEXT: kmovw %ebx, %k1 |
| ; AVX512F-32-NEXT: vmovss %xmm1, %xmm1, %xmm1 {%k1} {z} |
| ; AVX512F-32-NEXT: vsubss %xmm1, %xmm0, %xmm0 |
| ; AVX512F-32-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: flds (%esp) |
| ; AVX512F-32-NEXT: fisttpll (%esp) |
| ; AVX512F-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512F-32-NEXT: wait |
| ; AVX512F-32-NEXT: shll $31, %eax |
| ; AVX512F-32-NEXT: xorl {{[0-9]+}}(%esp), %eax |
| ; AVX512F-32-NEXT: shll $31, %edx |
| ; AVX512F-32-NEXT: xorl {{[0-9]+}}(%esp), %edx |
| ; AVX512F-32-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; AVX512F-32-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 |
| ; AVX512F-32-NEXT: vpinsrd $2, (%esp), %xmm0, %xmm0 |
| ; AVX512F-32-NEXT: vpinsrd $3, %eax, %xmm0, %xmm0 |
| ; AVX512F-32-NEXT: shll $31, %ecx |
| ; AVX512F-32-NEXT: xorl {{[0-9]+}}(%esp), %ecx |
| ; AVX512F-32-NEXT: shll $31, %ebx |
| ; AVX512F-32-NEXT: xorl {{[0-9]+}}(%esp), %ebx |
| ; AVX512F-32-NEXT: vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX512F-32-NEXT: vpinsrd $1, %ebx, %xmm1, %xmm1 |
| ; AVX512F-32-NEXT: vpinsrd $2, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX512F-32-NEXT: vpinsrd $3, %ecx, %xmm1, %xmm1 |
| ; AVX512F-32-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 |
| ; AVX512F-32-NEXT: leal -4(%ebp), %esp |
| ; AVX512F-32-NEXT: popl %ebx |
| ; AVX512F-32-NEXT: popl %ebp |
| ; AVX512F-32-NEXT: .cfi_def_cfa %esp, 4 |
| ; AVX512F-32-NEXT: retl |
| ; |
| ; AVX512F-64-LABEL: strict_vector_fptoui_v4f32_to_v4i64: |
| ; AVX512F-64: # %bb.0: |
| ; AVX512F-64-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3] |
| ; AVX512F-64-NEXT: vcvttss2usi %xmm1, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm1 |
| ; AVX512F-64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0] |
| ; AVX512F-64-NEXT: vcvttss2usi %xmm2, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0] |
| ; AVX512F-64-NEXT: vcvttss2usi %xmm0, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512F-64-NEXT: vmovshdup {{.*#+}} xmm0 = xmm0[1,1,3,3] |
| ; AVX512F-64-NEXT: vcvttss2usi %xmm0, %rax |
| ; AVX512F-64-NEXT: vmovq %rax, %xmm0 |
| ; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0] |
| ; AVX512F-64-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX512F-64-NEXT: retq |
| ; |
| ; AVX512VL-32-LABEL: strict_vector_fptoui_v4f32_to_v4i64: |
| ; AVX512VL-32: # %bb.0: |
| ; AVX512VL-32-NEXT: pushl %ebp |
| ; AVX512VL-32-NEXT: .cfi_def_cfa_offset 8 |
| ; AVX512VL-32-NEXT: .cfi_offset %ebp, -8 |
| ; AVX512VL-32-NEXT: movl %esp, %ebp |
| ; AVX512VL-32-NEXT: .cfi_def_cfa_register %ebp |
| ; AVX512VL-32-NEXT: pushl %ebx |
| ; AVX512VL-32-NEXT: andl $-8, %esp |
| ; AVX512VL-32-NEXT: subl $40, %esp |
| ; AVX512VL-32-NEXT: .cfi_offset %ebx, -12 |
| ; AVX512VL-32-NEXT: vpermilps {{.*#+}} xmm2 = xmm0[3,3,3,3] |
| ; AVX512VL-32-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX512VL-32-NEXT: xorl %eax, %eax |
| ; AVX512VL-32-NEXT: vcomiss %xmm1, %xmm2 |
| ; AVX512VL-32-NEXT: setae %al |
| ; AVX512VL-32-NEXT: kmovw %eax, %k1 |
| ; AVX512VL-32-NEXT: vmovss %xmm1, %xmm1, %xmm3 {%k1} {z} |
| ; AVX512VL-32-NEXT: vsubss %xmm3, %xmm2, %xmm2 |
| ; AVX512VL-32-NEXT: vmovss %xmm2, (%esp) |
| ; AVX512VL-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0] |
| ; AVX512VL-32-NEXT: xorl %edx, %edx |
| ; AVX512VL-32-NEXT: vcomiss %xmm1, %xmm2 |
| ; AVX512VL-32-NEXT: setae %dl |
| ; AVX512VL-32-NEXT: kmovw %edx, %k1 |
| ; AVX512VL-32-NEXT: vmovss %xmm1, %xmm1, %xmm3 {%k1} {z} |
| ; AVX512VL-32-NEXT: vsubss %xmm3, %xmm2, %xmm2 |
| ; AVX512VL-32-NEXT: vmovss %xmm2, {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3] |
| ; AVX512VL-32-NEXT: xorl %ecx, %ecx |
| ; AVX512VL-32-NEXT: vcomiss %xmm1, %xmm2 |
| ; AVX512VL-32-NEXT: setae %cl |
| ; AVX512VL-32-NEXT: kmovw %ecx, %k1 |
| ; AVX512VL-32-NEXT: vmovss %xmm1, %xmm1, %xmm3 {%k1} {z} |
| ; AVX512VL-32-NEXT: vsubss %xmm3, %xmm2, %xmm2 |
| ; AVX512VL-32-NEXT: vmovss %xmm2, {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: xorl %ebx, %ebx |
| ; AVX512VL-32-NEXT: vcomiss %xmm1, %xmm0 |
| ; AVX512VL-32-NEXT: setae %bl |
| ; AVX512VL-32-NEXT: kmovw %ebx, %k1 |
| ; AVX512VL-32-NEXT: vmovss %xmm1, %xmm1, %xmm1 {%k1} {z} |
| ; AVX512VL-32-NEXT: vsubss %xmm1, %xmm0, %xmm0 |
| ; AVX512VL-32-NEXT: vmovss %xmm0, {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: flds (%esp) |
| ; AVX512VL-32-NEXT: fisttpll (%esp) |
| ; AVX512VL-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: flds {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: fisttpll {{[0-9]+}}(%esp) |
| ; AVX512VL-32-NEXT: wait |
| ; AVX512VL-32-NEXT: shll $31, %eax |
| ; AVX512VL-32-NEXT: xorl {{[0-9]+}}(%esp), %eax |
| ; AVX512VL-32-NEXT: shll $31, %edx |
| ; AVX512VL-32-NEXT: xorl {{[0-9]+}}(%esp), %edx |
| ; AVX512VL-32-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero |
| ; AVX512VL-32-NEXT: vpinsrd $1, %edx, %xmm0, %xmm0 |
| ; AVX512VL-32-NEXT: vpinsrd $2, (%esp), %xmm0, %xmm0 |
| ; AVX512VL-32-NEXT: vpinsrd $3, %eax, %xmm0, %xmm0 |
| ; AVX512VL-32-NEXT: shll $31, %ecx |
| ; AVX512VL-32-NEXT: xorl {{[0-9]+}}(%esp), %ecx |
| ; AVX512VL-32-NEXT: shll $31, %ebx |
| ; AVX512VL-32-NEXT: xorl {{[0-9]+}}(%esp), %ebx |
| ; AVX512VL-32-NEXT: vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero |
| ; AVX512VL-32-NEXT: vpinsrd $1, %ebx, %xmm1, %xmm1 |
| ; AVX512VL-32-NEXT: vpinsrd $2, {{[0-9]+}}(%esp), %xmm1, %xmm1 |
| ; AVX512VL-32-NEXT: vpinsrd $3, %ecx, %xmm1, %xmm1 |
| ; AVX512VL-32-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 |
| ; AVX512VL-32-NEXT: leal -4(%ebp), %esp |
| ; AVX512VL-32-NEXT: popl %ebx |
| ; AVX512VL-32-NEXT: popl %ebp |
| ; AVX512VL-32-NEXT: .cfi_def_cfa %esp, 4 |
| ; AVX512VL-32-NEXT: retl |
| ; |
| ; AVX512VL-64-LABEL: strict_vector_fptoui_v4f32_to_v4i64: |
| ; AVX512VL-64: # %bb.0: |
| ; AVX512VL-64-NEXT: vpermilps {{.*#+}} xmm1 = xmm0[3,3,3,3] |
| ; AVX512VL-64-NEXT: vcvttss2usi %xmm1, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm1 |
| ; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0] |
| ; AVX512VL-64-NEXT: vcvttss2usi %xmm2, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0] |
| ; AVX512VL-64-NEXT: vcvttss2usi %xmm0, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm2 |
| ; AVX512VL-64-NEXT: vmovshdup {{.*#+}} xmm0 = xmm0[1,1,3,3] |
| ; AVX512VL-64-NEXT: vcvttss2usi %xmm0, %rax |
| ; AVX512VL-64-NEXT: vmovq %rax, %xmm0 |
| ; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0] |
| ; AVX512VL-64-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 |
| ; AVX512VL-64-NEXT: retq |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptoui_v4f32_to_v4i64: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vmovaps %xmm0, %xmm0 |
| ; AVX512DQ-NEXT: vcvttps2uqq %ymm0, %zmm0 |
| ; AVX512DQ-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0 |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptoui_v4f32_to_v4i64: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttps2uqq %xmm0, %ymm0 |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <4 x i64> @llvm.experimental.constrained.fptoui.v4i64.v4f32(<4 x float> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <4 x i64> %ret |
| } |
| |
| define <4 x i32> @strict_vector_fptosi_v4f64_to_v4i32(<4 x double> %a) #0 { |
| ; CHECK-LABEL: strict_vector_fptosi_v4f64_to_v4i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; CHECK-NEXT: vzeroupper |
| ; CHECK-NEXT: ret{{[l|q]}} |
| %ret = call <4 x i32> @llvm.experimental.constrained.fptosi.v4i32.v4f64(<4 x double> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <4 x i32> %ret |
| } |
| |
| define <4 x i32> @strict_vector_fptoui_v4f64_to_v4i32(<4 x double> %a) #0 { |
| ; AVX-LABEL: strict_vector_fptoui_v4f64_to_v4i32: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vmovapd {{.*#+}} ymm1 = [2.147483648E+9,2.147483648E+9,2.147483648E+9,2.147483648E+9] |
| ; AVX-NEXT: vcmpltpd %ymm1, %ymm0, %ymm2 |
| ; AVX-NEXT: vextractf128 $1, %ymm2, %xmm3 |
| ; AVX-NEXT: vshufps {{.*#+}} xmm3 = xmm2[0,2],xmm3[0,2] |
| ; AVX-NEXT: vxorps %xmm4, %xmm4, %xmm4 |
| ; AVX-NEXT: vmovaps {{.*#+}} xmm5 = [2147483648,2147483648,2147483648,2147483648] |
| ; AVX-NEXT: vblendvps %xmm3, %xmm4, %xmm5, %xmm3 |
| ; AVX-NEXT: vxorps %xmm4, %xmm4, %xmm4 |
| ; AVX-NEXT: vblendvpd %ymm2, %ymm4, %ymm1, %ymm1 |
| ; AVX-NEXT: vsubpd %ymm1, %ymm0, %ymm0 |
| ; AVX-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX-NEXT: vxorpd %xmm3, %xmm0, %xmm0 |
| ; AVX-NEXT: vzeroupper |
| ; AVX-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512F-LABEL: strict_vector_fptoui_v4f64_to_v4i32: |
| ; AVX512F: # %bb.0: |
| ; AVX512F-NEXT: vmovaps %ymm0, %ymm0 |
| ; AVX512F-NEXT: vcvttpd2udq %zmm0, %ymm0 |
| ; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 |
| ; AVX512F-NEXT: vzeroupper |
| ; AVX512F-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512VL-LABEL: strict_vector_fptoui_v4f64_to_v4i32: |
| ; AVX512VL: # %bb.0: |
| ; AVX512VL-NEXT: vcvttpd2udq %ymm0, %xmm0 |
| ; AVX512VL-NEXT: vzeroupper |
| ; AVX512VL-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptoui_v4f64_to_v4i32: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vmovaps %ymm0, %ymm0 |
| ; AVX512DQ-NEXT: vcvttpd2udq %zmm0, %ymm0 |
| ; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 |
| ; AVX512DQ-NEXT: vzeroupper |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptoui_v4f64_to_v4i32: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttpd2udq %ymm0, %xmm0 |
| ; AVX512DQVL-NEXT: vzeroupper |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <4 x i32> @llvm.experimental.constrained.fptoui.v4i32.v4f64(<4 x double> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <4 x i32> %ret |
| } |
| |
| define <4 x i16> @strict_vector_fptosi_v4f64_to_v4i16(<4 x double> %a) #0 { |
| ; CHECK-LABEL: strict_vector_fptosi_v4f64_to_v4i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; CHECK-NEXT: vpackssdw %xmm0, %xmm0, %xmm0 |
| ; CHECK-NEXT: vzeroupper |
| ; CHECK-NEXT: ret{{[l|q]}} |
| %ret = call <4 x i16> @llvm.experimental.constrained.fptosi.v4i16.v4f64(<4 x double> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <4 x i16> %ret |
| } |
| |
| define <4 x i16> @strict_vector_fptoui_v4f64_to_v4i16(<4 x double> %a) #0 { |
| ; CHECK-LABEL: strict_vector_fptoui_v4f64_to_v4i16: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; CHECK-NEXT: vpackusdw %xmm0, %xmm0, %xmm0 |
| ; CHECK-NEXT: vzeroupper |
| ; CHECK-NEXT: ret{{[l|q]}} |
| %ret = call <4 x i16> @llvm.experimental.constrained.fptoui.v4i16.v4f64(<4 x double> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <4 x i16> %ret |
| } |
| |
| define <4 x i8> @strict_vector_fptosi_v4f64_to_v4i8(<4 x double> %a) #0 { |
| ; AVX-LABEL: strict_vector_fptosi_v4f64_to_v4i8: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX-NEXT: vpackssdw %xmm0, %xmm0, %xmm0 |
| ; AVX-NEXT: vpacksswb %xmm0, %xmm0, %xmm0 |
| ; AVX-NEXT: vzeroupper |
| ; AVX-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512F-LABEL: strict_vector_fptosi_v4f64_to_v4i8: |
| ; AVX512F: # %bb.0: |
| ; AVX512F-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512F-NEXT: vpackssdw %xmm0, %xmm0, %xmm0 |
| ; AVX512F-NEXT: vpacksswb %xmm0, %xmm0, %xmm0 |
| ; AVX512F-NEXT: vzeroupper |
| ; AVX512F-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512VL-LABEL: strict_vector_fptosi_v4f64_to_v4i8: |
| ; AVX512VL: # %bb.0: |
| ; AVX512VL-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512VL-NEXT: vpmovdb %xmm0, %xmm0 |
| ; AVX512VL-NEXT: vzeroupper |
| ; AVX512VL-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptosi_v4f64_to_v4i8: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512DQ-NEXT: vpackssdw %xmm0, %xmm0, %xmm0 |
| ; AVX512DQ-NEXT: vpacksswb %xmm0, %xmm0, %xmm0 |
| ; AVX512DQ-NEXT: vzeroupper |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptosi_v4f64_to_v4i8: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512DQVL-NEXT: vpmovdb %xmm0, %xmm0 |
| ; AVX512DQVL-NEXT: vzeroupper |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <4 x i8> @llvm.experimental.constrained.fptosi.v4i8.v4f64(<4 x double> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <4 x i8> %ret |
| } |
| |
| define <4 x i8> @strict_vector_fptoui_v4f64_to_v4i8(<4 x double> %a) #0 { |
| ; AVX-LABEL: strict_vector_fptoui_v4f64_to_v4i8: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX-NEXT: vpackusdw %xmm0, %xmm0, %xmm0 |
| ; AVX-NEXT: vpackuswb %xmm0, %xmm0, %xmm0 |
| ; AVX-NEXT: vzeroupper |
| ; AVX-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512F-LABEL: strict_vector_fptoui_v4f64_to_v4i8: |
| ; AVX512F: # %bb.0: |
| ; AVX512F-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512F-NEXT: vpackusdw %xmm0, %xmm0, %xmm0 |
| ; AVX512F-NEXT: vpackuswb %xmm0, %xmm0, %xmm0 |
| ; AVX512F-NEXT: vzeroupper |
| ; AVX512F-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512VL-LABEL: strict_vector_fptoui_v4f64_to_v4i8: |
| ; AVX512VL: # %bb.0: |
| ; AVX512VL-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512VL-NEXT: vpmovdb %xmm0, %xmm0 |
| ; AVX512VL-NEXT: vzeroupper |
| ; AVX512VL-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptoui_v4f64_to_v4i8: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512DQ-NEXT: vpackusdw %xmm0, %xmm0, %xmm0 |
| ; AVX512DQ-NEXT: vpackuswb %xmm0, %xmm0, %xmm0 |
| ; AVX512DQ-NEXT: vzeroupper |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptoui_v4f64_to_v4i8: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512DQVL-NEXT: vpmovdb %xmm0, %xmm0 |
| ; AVX512DQVL-NEXT: vzeroupper |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <4 x i8> @llvm.experimental.constrained.fptoui.v4i8.v4f64(<4 x double> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <4 x i8> %ret |
| } |
| |
| define <4 x i1> @strict_vector_fptosi_v4f64_to_v4i1(<4 x double> %a) #0 { |
| ; AVX-LABEL: strict_vector_fptosi_v4f64_to_v4i1: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX-NEXT: vzeroupper |
| ; AVX-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512F-LABEL: strict_vector_fptosi_v4f64_to_v4i1: |
| ; AVX512F: # %bb.0: |
| ; AVX512F-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512F-NEXT: vptestmd %zmm0, %zmm0, %k1 |
| ; AVX512F-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z} |
| ; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0 |
| ; AVX512F-NEXT: vzeroupper |
| ; AVX512F-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512VL-LABEL: strict_vector_fptosi_v4f64_to_v4i1: |
| ; AVX512VL: # %bb.0: |
| ; AVX512VL-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512VL-NEXT: vptestmd %xmm0, %xmm0, %k1 |
| ; AVX512VL-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0 |
| ; AVX512VL-NEXT: vmovdqa32 %xmm0, %xmm0 {%k1} {z} |
| ; AVX512VL-NEXT: vzeroupper |
| ; AVX512VL-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptosi_v4f64_to_v4i1: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512DQ-NEXT: vpmovd2m %zmm0, %k0 |
| ; AVX512DQ-NEXT: vpmovm2d %k0, %zmm0 |
| ; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0 |
| ; AVX512DQ-NEXT: vzeroupper |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptosi_v4f64_to_v4i1: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512DQVL-NEXT: vpmovd2m %xmm0, %k0 |
| ; AVX512DQVL-NEXT: vpmovm2d %k0, %xmm0 |
| ; AVX512DQVL-NEXT: vzeroupper |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <4 x i1> @llvm.experimental.constrained.fptosi.v4i1.v4f64(<4 x double> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <4 x i1> %ret |
| } |
| |
| define <4 x i1> @strict_vector_fptoui_v4f64_to_v4i1(<4 x double> %a) #0 { |
| ; AVX-LABEL: strict_vector_fptoui_v4f64_to_v4i1: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX-NEXT: vzeroupper |
| ; AVX-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512F-LABEL: strict_vector_fptoui_v4f64_to_v4i1: |
| ; AVX512F: # %bb.0: |
| ; AVX512F-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512F-NEXT: vpslld $31, %xmm0, %xmm0 |
| ; AVX512F-NEXT: vptestmd %zmm0, %zmm0, %k1 |
| ; AVX512F-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z} |
| ; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0 |
| ; AVX512F-NEXT: vzeroupper |
| ; AVX512F-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512VL-LABEL: strict_vector_fptoui_v4f64_to_v4i1: |
| ; AVX512VL: # %bb.0: |
| ; AVX512VL-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512VL-NEXT: vpslld $31, %xmm0, %xmm0 |
| ; AVX512VL-NEXT: vptestmd %xmm0, %xmm0, %k1 |
| ; AVX512VL-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0 |
| ; AVX512VL-NEXT: vmovdqa32 %xmm0, %xmm0 {%k1} {z} |
| ; AVX512VL-NEXT: vzeroupper |
| ; AVX512VL-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptoui_v4f64_to_v4i1: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512DQ-NEXT: vpslld $31, %xmm0, %xmm0 |
| ; AVX512DQ-NEXT: vpmovd2m %zmm0, %k0 |
| ; AVX512DQ-NEXT: vpmovm2d %k0, %zmm0 |
| ; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0 |
| ; AVX512DQ-NEXT: vzeroupper |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptoui_v4f64_to_v4i1: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttpd2dq %ymm0, %xmm0 |
| ; AVX512DQVL-NEXT: vpslld $31, %xmm0, %xmm0 |
| ; AVX512DQVL-NEXT: vpmovd2m %xmm0, %k0 |
| ; AVX512DQVL-NEXT: vpmovm2d %k0, %xmm0 |
| ; AVX512DQVL-NEXT: vzeroupper |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <4 x i1> @llvm.experimental.constrained.fptoui.v4i1.v4f64(<4 x double> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <4 x i1> %ret |
| } |
| |
| define <8 x i32> @strict_vector_fptosi_v8f32_to_v8i32(<8 x float> %a) #0 { |
| ; CHECK-LABEL: strict_vector_fptosi_v8f32_to_v8i32: |
| ; CHECK: # %bb.0: |
| ; CHECK-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; CHECK-NEXT: ret{{[l|q]}} |
| %ret = call <8 x i32> @llvm.experimental.constrained.fptosi.v8i32.v8f32(<8 x float> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <8 x i32> %ret |
| } |
| |
| define <8 x i32> @strict_vector_fptoui_v8f32_to_v8i32(<8 x float> %a) #0 { |
| ; AVX-LABEL: strict_vector_fptoui_v8f32_to_v8i32: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vmovaps {{.*#+}} ymm1 = [2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9,2.14748365E+9] |
| ; AVX-NEXT: vcmpltps %ymm1, %ymm0, %ymm2 |
| ; AVX-NEXT: vxorps %xmm3, %xmm3, %xmm3 |
| ; AVX-NEXT: vmovaps {{.*#+}} ymm4 = [2147483648,2147483648,2147483648,2147483648,2147483648,2147483648,2147483648,2147483648] |
| ; AVX-NEXT: vblendvps %ymm2, %ymm3, %ymm4, %ymm4 |
| ; AVX-NEXT: vblendvps %ymm2, %ymm3, %ymm1, %ymm1 |
| ; AVX-NEXT: vsubps %ymm1, %ymm0, %ymm0 |
| ; AVX-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX-NEXT: vxorps %ymm4, %ymm0, %ymm0 |
| ; AVX-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512F-LABEL: strict_vector_fptoui_v8f32_to_v8i32: |
| ; AVX512F: # %bb.0: |
| ; AVX512F-NEXT: vmovaps %ymm0, %ymm0 |
| ; AVX512F-NEXT: vcvttps2udq %zmm0, %zmm0 |
| ; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0 |
| ; AVX512F-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512VL-LABEL: strict_vector_fptoui_v8f32_to_v8i32: |
| ; AVX512VL: # %bb.0: |
| ; AVX512VL-NEXT: vcvttps2udq %ymm0, %ymm0 |
| ; AVX512VL-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptoui_v8f32_to_v8i32: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vmovaps %ymm0, %ymm0 |
| ; AVX512DQ-NEXT: vcvttps2udq %zmm0, %zmm0 |
| ; AVX512DQ-NEXT: # kill: def $ymm0 killed $ymm0 killed $zmm0 |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptoui_v8f32_to_v8i32: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttps2udq %ymm0, %ymm0 |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <8 x i32> @llvm.experimental.constrained.fptoui.v8i32.v8f32(<8 x float> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <8 x i32> %ret |
| } |
| |
| define <8 x i16> @strict_vector_fptosi_v8f32_to_v8i16(<8 x float> %a) #0 { |
| ; AVX-LABEL: strict_vector_fptosi_v8f32_to_v8i16: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1 |
| ; AVX-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 |
| ; AVX-NEXT: vzeroupper |
| ; AVX-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512F-LABEL: strict_vector_fptosi_v8f32_to_v8i16: |
| ; AVX512F: # %bb.0: |
| ; AVX512F-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512F-NEXT: vpmovdw %zmm0, %ymm0 |
| ; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 |
| ; AVX512F-NEXT: vzeroupper |
| ; AVX512F-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512VL-LABEL: strict_vector_fptosi_v8f32_to_v8i16: |
| ; AVX512VL: # %bb.0: |
| ; AVX512VL-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512VL-NEXT: vpmovdw %ymm0, %xmm0 |
| ; AVX512VL-NEXT: vzeroupper |
| ; AVX512VL-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptosi_v8f32_to_v8i16: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512DQ-NEXT: vpmovdw %zmm0, %ymm0 |
| ; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 |
| ; AVX512DQ-NEXT: vzeroupper |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptosi_v8f32_to_v8i16: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512DQVL-NEXT: vpmovdw %ymm0, %xmm0 |
| ; AVX512DQVL-NEXT: vzeroupper |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <8 x i16> @llvm.experimental.constrained.fptosi.v8i16.v8f32(<8 x float> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <8 x i16> %ret |
| } |
| |
| define <8 x i16> @strict_vector_fptoui_v8f32_to_v8i16(<8 x float> %a) #0 { |
| ; AVX-LABEL: strict_vector_fptoui_v8f32_to_v8i16: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1 |
| ; AVX-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 |
| ; AVX-NEXT: vzeroupper |
| ; AVX-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512F-LABEL: strict_vector_fptoui_v8f32_to_v8i16: |
| ; AVX512F: # %bb.0: |
| ; AVX512F-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512F-NEXT: vpmovdw %zmm0, %ymm0 |
| ; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 |
| ; AVX512F-NEXT: vzeroupper |
| ; AVX512F-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512VL-LABEL: strict_vector_fptoui_v8f32_to_v8i16: |
| ; AVX512VL: # %bb.0: |
| ; AVX512VL-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512VL-NEXT: vpmovdw %ymm0, %xmm0 |
| ; AVX512VL-NEXT: vzeroupper |
| ; AVX512VL-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptoui_v8f32_to_v8i16: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512DQ-NEXT: vpmovdw %zmm0, %ymm0 |
| ; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 |
| ; AVX512DQ-NEXT: vzeroupper |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptoui_v8f32_to_v8i16: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512DQVL-NEXT: vpmovdw %ymm0, %xmm0 |
| ; AVX512DQVL-NEXT: vzeroupper |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <8 x i16> @llvm.experimental.constrained.fptoui.v8i16.v8f32(<8 x float> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <8 x i16> %ret |
| } |
| |
| define <8 x i8> @strict_vector_fptosi_v8f32_to_v8i8(<8 x float> %a) #0 { |
| ; AVX-LABEL: strict_vector_fptosi_v8f32_to_v8i8: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1 |
| ; AVX-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 |
| ; AVX-NEXT: vpacksswb %xmm0, %xmm0, %xmm0 |
| ; AVX-NEXT: vzeroupper |
| ; AVX-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512F-LABEL: strict_vector_fptosi_v8f32_to_v8i8: |
| ; AVX512F: # %bb.0: |
| ; AVX512F-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512F-NEXT: vpmovdb %zmm0, %xmm0 |
| ; AVX512F-NEXT: vzeroupper |
| ; AVX512F-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512VL-LABEL: strict_vector_fptosi_v8f32_to_v8i8: |
| ; AVX512VL: # %bb.0: |
| ; AVX512VL-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512VL-NEXT: vpmovdb %ymm0, %xmm0 |
| ; AVX512VL-NEXT: vzeroupper |
| ; AVX512VL-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptosi_v8f32_to_v8i8: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512DQ-NEXT: vpmovdb %zmm0, %xmm0 |
| ; AVX512DQ-NEXT: vzeroupper |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptosi_v8f32_to_v8i8: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512DQVL-NEXT: vpmovdb %ymm0, %xmm0 |
| ; AVX512DQVL-NEXT: vzeroupper |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <8 x i8> @llvm.experimental.constrained.fptosi.v8i8.v8f32(<8 x float> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <8 x i8> %ret |
| } |
| |
| define <8 x i8> @strict_vector_fptoui_v8f32_to_v8i8(<8 x float> %a) #0 { |
| ; AVX-LABEL: strict_vector_fptoui_v8f32_to_v8i8: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1 |
| ; AVX-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 |
| ; AVX-NEXT: vpackuswb %xmm0, %xmm0, %xmm0 |
| ; AVX-NEXT: vzeroupper |
| ; AVX-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512F-LABEL: strict_vector_fptoui_v8f32_to_v8i8: |
| ; AVX512F: # %bb.0: |
| ; AVX512F-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512F-NEXT: vpmovdb %zmm0, %xmm0 |
| ; AVX512F-NEXT: vzeroupper |
| ; AVX512F-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512VL-LABEL: strict_vector_fptoui_v8f32_to_v8i8: |
| ; AVX512VL: # %bb.0: |
| ; AVX512VL-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512VL-NEXT: vpmovdb %ymm0, %xmm0 |
| ; AVX512VL-NEXT: vzeroupper |
| ; AVX512VL-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptoui_v8f32_to_v8i8: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512DQ-NEXT: vpmovdb %zmm0, %xmm0 |
| ; AVX512DQ-NEXT: vzeroupper |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptoui_v8f32_to_v8i8: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512DQVL-NEXT: vpmovdb %ymm0, %xmm0 |
| ; AVX512DQVL-NEXT: vzeroupper |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <8 x i8> @llvm.experimental.constrained.fptoui.v8i8.v8f32(<8 x float> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <8 x i8> %ret |
| } |
| |
| define <8 x i1> @strict_vector_fptosi_v8f32_to_v8i1(<8 x float> %a) #0 { |
| ; AVX-LABEL: strict_vector_fptosi_v8f32_to_v8i1: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1 |
| ; AVX-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 |
| ; AVX-NEXT: vzeroupper |
| ; AVX-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512F-LABEL: strict_vector_fptosi_v8f32_to_v8i1: |
| ; AVX512F: # %bb.0: |
| ; AVX512F-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512F-NEXT: vptestmd %zmm0, %zmm0, %k1 |
| ; AVX512F-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z} |
| ; AVX512F-NEXT: vpmovdw %zmm0, %ymm0 |
| ; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 |
| ; AVX512F-NEXT: vzeroupper |
| ; AVX512F-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512VL-LABEL: strict_vector_fptosi_v8f32_to_v8i1: |
| ; AVX512VL: # %bb.0: |
| ; AVX512VL-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512VL-NEXT: vptestmd %ymm0, %ymm0, %k1 |
| ; AVX512VL-NEXT: vpcmpeqd %ymm0, %ymm0, %ymm0 |
| ; AVX512VL-NEXT: vmovdqa32 %ymm0, %ymm0 {%k1} {z} |
| ; AVX512VL-NEXT: vpmovdw %ymm0, %xmm0 |
| ; AVX512VL-NEXT: vzeroupper |
| ; AVX512VL-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptosi_v8f32_to_v8i1: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512DQ-NEXT: vpmovd2m %zmm0, %k0 |
| ; AVX512DQ-NEXT: vpmovm2d %k0, %zmm0 |
| ; AVX512DQ-NEXT: vpmovdw %zmm0, %ymm0 |
| ; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 |
| ; AVX512DQ-NEXT: vzeroupper |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptosi_v8f32_to_v8i1: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512DQVL-NEXT: vpmovd2m %ymm0, %k0 |
| ; AVX512DQVL-NEXT: vpmovm2d %k0, %ymm0 |
| ; AVX512DQVL-NEXT: vpmovdw %ymm0, %xmm0 |
| ; AVX512DQVL-NEXT: vzeroupper |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <8 x i1> @llvm.experimental.constrained.fptosi.v8i1.v8f32(<8 x float> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <8 x i1> %ret |
| } |
| |
| define <8 x i1> @strict_vector_fptoui_v8f32_to_v8i1(<8 x float> %a) #0 { |
| ; AVX-LABEL: strict_vector_fptoui_v8f32_to_v8i1: |
| ; AVX: # %bb.0: |
| ; AVX-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1 |
| ; AVX-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 |
| ; AVX-NEXT: vzeroupper |
| ; AVX-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512F-LABEL: strict_vector_fptoui_v8f32_to_v8i1: |
| ; AVX512F: # %bb.0: |
| ; AVX512F-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512F-NEXT: vpslld $31, %ymm0, %ymm0 |
| ; AVX512F-NEXT: vptestmd %zmm0, %zmm0, %k1 |
| ; AVX512F-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z} |
| ; AVX512F-NEXT: vpmovdw %zmm0, %ymm0 |
| ; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 |
| ; AVX512F-NEXT: vzeroupper |
| ; AVX512F-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512VL-LABEL: strict_vector_fptoui_v8f32_to_v8i1: |
| ; AVX512VL: # %bb.0: |
| ; AVX512VL-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512VL-NEXT: vpslld $31, %ymm0, %ymm0 |
| ; AVX512VL-NEXT: vptestmd %ymm0, %ymm0, %k1 |
| ; AVX512VL-NEXT: vpcmpeqd %ymm0, %ymm0, %ymm0 |
| ; AVX512VL-NEXT: vmovdqa32 %ymm0, %ymm0 {%k1} {z} |
| ; AVX512VL-NEXT: vpmovdw %ymm0, %xmm0 |
| ; AVX512VL-NEXT: vzeroupper |
| ; AVX512VL-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQ-LABEL: strict_vector_fptoui_v8f32_to_v8i1: |
| ; AVX512DQ: # %bb.0: |
| ; AVX512DQ-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512DQ-NEXT: vpslld $31, %ymm0, %ymm0 |
| ; AVX512DQ-NEXT: vpmovd2m %zmm0, %k0 |
| ; AVX512DQ-NEXT: vpmovm2d %k0, %zmm0 |
| ; AVX512DQ-NEXT: vpmovdw %zmm0, %ymm0 |
| ; AVX512DQ-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 |
| ; AVX512DQ-NEXT: vzeroupper |
| ; AVX512DQ-NEXT: ret{{[l|q]}} |
| ; |
| ; AVX512DQVL-LABEL: strict_vector_fptoui_v8f32_to_v8i1: |
| ; AVX512DQVL: # %bb.0: |
| ; AVX512DQVL-NEXT: vcvttps2dq %ymm0, %ymm0 |
| ; AVX512DQVL-NEXT: vpslld $31, %ymm0, %ymm0 |
| ; AVX512DQVL-NEXT: vpmovd2m %ymm0, %k0 |
| ; AVX512DQVL-NEXT: vpmovm2d %k0, %ymm0 |
| ; AVX512DQVL-NEXT: vpmovdw %ymm0, %xmm0 |
| ; AVX512DQVL-NEXT: vzeroupper |
| ; AVX512DQVL-NEXT: ret{{[l|q]}} |
| %ret = call <8 x i1> @llvm.experimental.constrained.fptoui.v8i1.v8f32(<8 x float> %a, |
| metadata !"fpexcept.strict") #0 |
| ret <8 x i1> %ret |
| } |
| |
| attributes #0 = { strictfp } |